Predictive data analysis using image representations of categorical and scalar feature data

Abstract

There is a need for more effective and efficient predictive data analysis solutions and/or more effective and efficient solutions for generating image representations of categorical/scalar data. Various embodiments of the present invention address one or more of the noted technical challenges. In one example, a method comprises receiving the one or more categorical input features; generating an image representation of the one or more categorical input features, wherein the image representation comprises image region values each associated with a categorical input feature, and further wherein each image region value of the one or more image region values is determined based at least in part on the corresponding categorical input feature associated with the image region value; and processing the image representation using an image-based machine learning model to generate the image-based predictions.

The invention claimed is: 
     
       1. A computer-implemented method comprising:
 receiving, by one or more processors, one or more categorical input features; 
 determining, by the one or more processors, a plurality of image region values, wherein each image region value of the plurality of image region values is (i) associated with a respective image region of a plurality of image regions, (ii) i-s associated with a respective categorical input feature type of one or more categorical input feature types, and (iii) determined in a manner such that the respective image region for the image region value displays a visual representation of a respective categorical input feature for the respective categorical input feature type that is associated with the respective image region for the image region value; 
 generating, by the one or more processors, an image representation based at least in part on one or more image region values corresponding to a categorical input feature of the one or more categorical input features; and 
 generating, by the one or more processors and using an image-based machine learning model, and based at least in part on the image representation, an image-based prediction. 
 
     
     
       2. The computer-implemented method of  claim 1 , wherein:
 the one or more categorical input features comprise one or more patient features associated with a patient, and 
 the image-based prediction is a health prediction for the patient. 
 
     
     
       3. The computer-implemented method of  claim 1 , wherein the computer-implemented method is performed in response to selecting a first image generation technique of a plurality of image generation techniques. 
     
     
       4. The computer-implemented method of  claim 3 , wherein:
 the plurality of image generation techniques comprises a second image generation technique, and 
 the second image generation technique comprises:
 identifying a plurality of character patterns; 
 generating, for each character pattern of the plurality of character patterns, a feature-based channel of a plurality of feature-based channels, wherein: (i) each feature-based channel comprises one or more feature-based channel region values, and (ii) each feature-based channel region value for a corresponding feature-based channel is associated with a corresponding categorical input feature, and (iii) each feature-based channel region value for a corresponding feature-based channel is determined based at least in part on whether the corresponding categorical input feature for the feature-based channel region value comprises the corresponding character pattern associated with the corresponding feature-based channel; and 
 generating the image representation based at least in part on one or more image region values corresponding to a categorical input feature of the one or more categorical input features. 
 
 
     
     
       5. The computer-implemented method of  claim 3 , wherein:
 the plurality of image generation techniques comprises a third image generation technique, and 
 the third image generation technique comprises:
 determining, for each categorical input feature of the one or more categorical input features, a corresponding coordinate grouping of a plurality of coordinate groupings; 
 generating a plurality of coordinate channels by generating a coordinate channel for each coordinate grouping of the plurality of coordinate groupings; and 
 generating the image representation based at least in part on one or more coordinate channels of the plurality of coordinate channels. 
 
 
     
     
       6. The computer-implemented method of  claim 3 , wherein:
 the plurality of image generation techniques comprises a fourth image generation technique, and 
 the fourth image generation technique comprises:
 generating, based at least in part on one or more categorical input feature values, one or more coordinate channels and one or more feature-based channels; and 
 merging the one or more coordinate channels and the one or more feature-based channels to generate the image representation. 
 
 
     
     
       7. The computer-implemented method of  claim 1 , wherein the image-based machine learning model comprises a convolutional neural network (CNN). 
     
     
       8. A system comprising memory and one or more processors configured to:
 receive one or more categorical input features; 
 determine a plurality of image region values, wherein each image region value of the plurality of image region values is (i) associated with a respective image region of a plurality of image regions, (ii) associated with a respective categorical input feature type of one or more categorical input feature types, and (iii) determined in a manner such that the respective image region for the image region value displays a visual representation of a respective categorical input feature for the respective categorical input feature type that is associated with the respective image region for the image region value; 
 generate an image representation based at least in part on one or more image region values corresponding to a categorical input feature of the one or more categorical input features; and 
 generate, using an image-based machine learning model, and based at least in part on the image representation, an image-based prediction. 
 
     
     
       9. The system of  claim 8 , wherein:
 the one or more categorical input features comprise one or more patient features associated with a patient, and 
 the image-based prediction is a health prediction for the patient. 
 
     
     
       10. The system of  claim 8 , wherein the one or more processors are configured to generate the image-based prediction responsive to a selection of a first image generation technique of a plurality of image generation techniques. 
     
     
       11. The system of  claim 10 , wherein:
 the plurality of image generation techniques comprises a second image generation technique, and 
 the second image generation technique comprises:
 identifying a plurality of character patterns; 
 generating, for each character pattern of the plurality of character patterns, a feature-based channel of a plurality of feature-based channels, wherein: (i) each feature-based channel comprises one or more feature-based channel region values, and (ii) each feature-based channel region value for a corresponding feature-based channel is associated with a corresponding categorical input feature, and (iii) each feature-based channel region value for a corresponding feature-based channel is determined based at least in part on whether the corresponding categorical input feature for the feature-based channel region value comprises the corresponding character pattern associated with the corresponding feature-based channel; and 
 generating the image representation based at least in part on one or more image region values corresponding to a categorical input feature of the one or more categorical input features. 
 
 
     
     
       12. The system of  claim 10 , wherein:
 the plurality of image generation techniques comprises a third image generation technique, and 
 the third image generation technique comprises:
 determining, for each categorical input feature of the one or more categorical input features, a corresponding coordinate grouping of a plurality of coordinate groupings; 
 generating a plurality of coordinate channels by generating a coordinate channel for each coordinate grouping of the plurality of coordinate groupings; and 
 generating the image representation based at least in part on one or more coordinate channels of the plurality of coordinate channels. 
 
 
     
     
       13. The system of  claim 10 , wherein:
 the plurality of image generation techniques comprises a fourth image generation technique, and 
 the fourth image generation technique comprises:
 generating, based at least in part on one or more categorical input feature values, one or more coordinate channels and one or more feature-based channels; and 
 merging the one or more coordinate channels and the one or more feature-based channels to generate the image representation. 
 
 
     
     
       14. The system of  claim 8 , wherein the image-based machine learning model comprises a convolutional neural network (CNN). 
     
     
       15. One or more non-transitory computer readable storage media including instructions that cause one or more processors to:
 receive one or more categorical input features; 
 determine a plurality of image region values, wherein each image region value of the plurality of image region values is (i) associated with a respective image region of a plurality of image regions, (ii) associated with a respective categorical input feature type of one or more categorical input feature types, and (iii) determined in a manner such that the respective image region for the image region value displays a visual representation of a respective categorical input feature for the respective categorical input feature type that is associated with the respective image region for the image region value; 
 generate an image representation based at least in part on one or more image region values corresponding to a categorical input feature of the one or more categorical input features; and 
 generate, using an image-based machine learning model, and based at least in part on the image representation, an image-based prediction. 
 
     
     
       16. The one or more non-transitory computer readable storage media of  claim 15 , wherein:
 the one or more categorical input features comprise one or more patient features associated with a patient, and 
 the image-based prediction is a health prediction for the patient. 
 
     
     
       17. The one or more non-transitory computer readable storage media of  claim 15 , wherein one or more processors are caused to generate the image-based prediction responsive to a selection of a first image generation technique of a plurality of image generation techniques. 
     
     
       18. The one or more non-transitory computer readable storage media of  claim 17 , wherein:
 the plurality of image generation techniques comprises a second image generation technique, and 
 the second image generation technique comprises:
 identifying a plurality of character patterns; 
 generating, for each character pattern of the plurality of character patterns, a feature-based channel of a plurality of feature-based channels, wherein: (i) each feature-based channel comprises one or more feature-based channel region values, (ii) each feature-based channel region value for a corresponding feature-based channel is associated with a corresponding categorical input feature, and (iii) each feature-based channel region value for a corresponding feature-based channel is determined based at least in part on whether the corresponding categorical input feature for the feature-based channel region value comprises the corresponding character pattern associated with the corresponding feature-based channel; and 
 generating the image representation based at least in part on one or more image region values corresponding to a categorical input feature of the one or more categorical input features. 
 
 
     
     
       19. The one or more non-transitory computer readable storage media of  claim 17 , wherein:
 the plurality of image generation techniques comprises a third image generation technique, and 
 the third image generation technique comprises:
 determining, for each categorical input feature of the one or more categorical input features, a corresponding coordinate grouping of a plurality of coordinate groupings; 
 generating a plurality of coordinate channels by generating a coordinate channel for each coordinate grouping of the plurality of coordinate groupings; and 
 generating the image representation based at least in part on one or more coordinate channels of the plurality of coordinate channels. 
 
 
     
     
       20. The one or more non-transitory computer readable storage media of  claim 17 , wherein:
 the plurality of image generation techniques comprises a fourth image generation technique, and 
 the fourth image generation technique comprises:
 generating, based at least in part on one or more categorical input feature values, one or more coordinate channels and one or more feature-based channels; and 
 merging the one or more coordinate channels and the one or more feature-based channels to generate the image representation.